You have a database of 950 photos of cats and 50 photos of dogs. You have a program that tells cats from dogs automatically. You run it using your database, and you see that it reports 6% error rate. Does the program work well?

You have a database of 950 photos of cats and 50 photos of dogs. You have a program that tells cats from dogs automatically. You run it using your database, and you see that it reports 6% error rate. Does the program work well?

You have a database of 950 photos of cats and 50 photos of dogs. You have a program that tells cats from dogs automatically. You run it using your database, and you see that it reports 6% error rate. Does the program work well?



  1.     Yes, because it gives me a correct answer in 94% of cases.
  2.     Yes, because the probability to encounter an image picturing a cat is the same as that of a dog, and the program is only mistaked in 6% of the cases.
  3.     No, because it performs worse than a guess based on the distribution of dog and cat images in the database.
  4.     No, because it should give a maximum error of 1%.